The origin is an important factor affecting the quality of Chinese herbal medicine. The difference of origin leads to the uneven quality of Chinese herbal medicine. In order to maintain the market order, it is necessary to establish the method of identification of the origin of Chinese herbal medicine to identify and analyze the quality of Chinese herbal medicine more accurately. This article takes danshen, a major clinical medicinal material, from many origins as the research object, and 150 samples of danshen were collected from different origins. The surface of each root of danshen sample was scanned 1~n times randomly by micro focusing Raman spectroscopy under non-destructive conditions, and the average spectrum of each sample was calculated. By analyzing the original spectral data, it is found that the surface spectral signal of danshen contains both the Raman spectra of tanshinones and the fluorescence spectra of impurities. Mainly reflected in that danshen from different origins has their aggregation regions, and the signal intensity of the surface spectral signal of danshen is significantly weaker or stronger than that of tanshinones in the specific wavelength range. After preprocessing the spectral data of 1~n scans, the classification model of danshen origin was established by partial least squares discriminant analysis (partial least squares-discriminant analysis, PLS-DA) and random forest classification algorithm [no screening (random forest, RF) or screening of important variables (RF-VS)]. Results the training set and test set accuracy of the optimal model obtained by random 1 scanning were 88% and 87% respectively, and the samples with low quality and high quality could be distinguished with an accurate of 97%; the training set and test set accuracy of the optimal model obtained by random scanning 2 and 3 times were both 89% and 87% respectively. Combined with the operation efficiency of the model The spectrum obtained by random 1 scanning was selected, and the identification model of the origin of danshen was obtained by first derivative (1ST-D) pretreatment and RF-VS calculation. In conclusion, the micro focused Raman spectroscopy technology can establish a rapid and accurate prediction model of the origin of danshen under non-invasive conditions and provide a reference for the further application of this technology in identifying the origin and authenticity of expensive and scarce Chinese herbal medicine.